Microbial profileOverall sketch of microbial profile in the hospital indoor environments

From the 16S rRNA gene amplicon sequencing, the raw data consisted of 598,1256 reads (2.99 Gbps) for all samples, and the quality-filtered (Q30) data consisted of 591,4227 reads with an average value of 92,410 reads per sample, corresponding to 1439 different bacteria types after filtering and quality control. The mapped level of each taxonomic level, namely phyla, class, order, family, genus, and species, against the Silva 138.1 database (Quast et al. 2013) was 100%, 99.9%, 99.7%, 98.8%, 82.8%, and 9.98%, respectively. The limited percentage of identified bacterial species could be attributed to the short reads length that may not provide sufficient resolution for species identification. The rarefaction analysis conducted at the genus level demonstrates that the data obtained was adequate for taxonomic analysis. Despite the limitations of short reads, the analysis is able to provide reliable insights into the composition and diversity of bacterial genera within the sample (Figure S1).

At the genus level, there were 684 genera detected in total. Out of the 684 genera, a subset of 215 genera were identified as being particularly abundant, accounting for over 0.01% of the total 16S rRNA gene sequences at the genus level. These 215 genera were considered representative of the bacterial community and were selected for further analysis in subsequent steps. In terms of the dominating genera, Dechloromonas, Pseudomonas and Flavobacterium, Limnohabitans, Ralstonia, and Acineotobactor took the leading places (Figure S2a), taking 11.0%, 7.95%, 7.56%, 6.41%, 5.53%, and 4.17% of the community respectively. As Pseudomonas is the most concerned waterborne pathogen in healthcare facilities responsible for a wide spectrum of infections (such as pneumonia and urinary tract infections) in humans that can be associated with significant morbidity and mortality (Bonadonna et al. 2017), Dechloromonas occurs frequently in the soil and wastewater treatment systems linked with nitrogen cycling roles (Petriglieri et al. 2021) and Flavobacterium exists more in soil and freshwater that may cause disease in freshwater fish (Bernardet et al. 1996). Moreover, the pathogenic Acineotobactor is particularly noteworthy due to the potential implications of its species (e.g., A. baumannii, Acinetobacter nosocomialis, Acinetobacter pittii, and Acinetobacter seifertii) as a source of infections (such as pneumonia, meningitis, and bacteremia) in debilitated patients within the hospital (Harding et al. 2018). Apart from the dominating concerned genera, several other genera of interest were observed in the lower abundance in the studied environment. These included Enterobacter (0.806%), Stenotrophomonas (0.743%), Bacillus (0.418%), Staphylococcus (0.263%), Klebsiella (0.217%), Mycobacterium (0.561%), Streptococcus (0.116%), Escherichia-Shigella (0.0981%), and Neisseria (0.0116%). Among them, Staphylococcus, Klebsiella, Acinetobacter, and Enterobacter corresponded to the genus types of ESKAPE pathogens (Enterococcus faecium, S. aureus, K. pneumoniae, A. baumannii, Pseudomonas aeruginosa, and Enterobacter spp.). However, it is important to note that not all ESKAPE pathogens were detected in the hospital environment under study. For instance, Enterococcus genus was not classified. Although Enterococci are commonly associated with hospital-acquired infections and are notorious for their resistance to vancomycin (Brinkwirth et al. 2021; Hammerum et al. 2024), the limitations of this study, such as DNA extraction efficiency and the sensitivity of V4–V5 16S rRNA amplicon sequencing, may have hindered the identification of Enterococcus in the samples.

From species-level respect, the results of predominating species (Figure S2b) were concerning, as they indicated the presence of certain pathogens in the studied environments. Among the major species, Pseudomonas sp. was found to be the most prevalent, accounting for 45.4% of the samples. This pathogen showed a particularly significant emergence in the emergency and otolaryngology departments. Another alarming pathogen was the presence of Stenotrophomonas maltophilia, which accounted for 5.12%. This pathogen was most frequently detected in the pneumology department, indicating a potential association with respiratory-related conditions. Furthermore, the study identified the presence of K. pneumoniae in 1.80% of the samples. Although the proportion might be relatively lower compared to other pathogens, the detection of K. pneumoniae was still significant due to its potential to cause serious infections and its association with antibiotic resistance (Madebo et al. 2022). In the study conducted by Madebo et al. (2022), P. aeruginosa and K. pneumoniae were reported as the leading contaminants in the hospital as well. However, another dominating pathogen of S. aureus did not appear in our study which could potentially be attributed to a low mapping rate at the species level during the analysis.

Seasonal changes shaped the bacterial composition in each department

When examining microbial alpha diversity in the indoor hospital environments, a significant fluctuation was observed across different departments, as depicted in Fig. 2f. The otolaryngology and surgery departments displayed the highest microbial diversity, with Shannon index values of 2.80 and 2.66, respectively. In contrast, the pneumology department exhibited the lower microbial diversity, recording the lowest Shannon index value of 1.99. To further explain such variation of alpha diversity, the Shannon index values between air dust and water within specific department were compared. Interestingly, in the otolaryngology, surgery, and pneumology departments, air dust consistently displayed higher bacterial alpha diversity. The Shannon index values for air dust were 3.23, 3.43, and 2.50, respectively. The water medium exhibited slightly lower Shannon index values, measuring 2.70, 2.51, and 1.61, respectively. This finding aligns with the understanding that aerosols released within the hospital environment can serve as a reservoir for a diverse range of microorganisms, contributing to the observed higher alpha diversity in air dust samples. Previous studies, such as the one by Bonadonna et al. (2017) have also highlighted the presence of microorganisms in aerosols released through various routes within healthcare facilities. Apparently, the notable increase in alpha diversity within the respiratory-related department, specifically the pneumology department, gives rise to apprehension. This heightened diversity had the potential to introduce a wide range of bacteria, including pathogens, into the respiratory system, thereby elevating the risk of respiratory infections and exacerbating pre-existing respiratory conditions. Notably, aerosol contact has been established as a significant pathway for respiratory infections, such as tuberculosis or Legionnaires’ disease (Tang et al. 2006), wherein inhaled bacteria colonize the respiratory tract. Adding to the concern, the formation of bacterial biofilms within the respiratory tract poses a considerable threat. Aerosols can facilitate the dissemination of these biofilms, thereby promoting the transmission of antibiotic-resistant bacteria among individuals (Ding et al. 2023).

Fig. 2

a PCA clusters in the studied hospital indoor environment separated by early summer, late summer and winter, of which the ellipses were drawn in 95% confidence level. b PCA characteristics of two mediums. Due to the absence of air dust samples in winter, the analysis in water and air dust was conducted without including the water samples from winter. c PCA characteristics of seven departments. d Bacterial (genus-level) Shannon diversity in seven departments, where the inside line in the each box represents the median value. e, f Five key genera that contributed significantly to the observed community compositional differentiation between early summer and late summer in the overall hospital indoor environment

In order to assess the primary environmental factor influencing the microbial composition in the hospital indoor environments, a PCA was conducted, considering variations in seasons, mediums, and departments. The PCA results reveal the presence of three distinct clusters along the PC1 and PC2 axes, primarily separated by seasonal differences, rather than variations in departments and mediums (Fig. 2a, d, e). This suggests that the microbial structure within the studied indoor hospital environment significantly differed between summer and winter. The influence of seasons was further supported by the ANOSIM result (Table S5). To gain a deeper understanding of the clusters within the summer season, the summer samples were divided into two subgroups: early summer and late summer. Therefore, another round of PCA analysis was conducted specifically for these subgroups (Fig. 2a). The results uncover that the two summer clusters were composed of different genera components, with early summer and late summer exhibiting distinct microbial compositions. This differentiation within the summer season was strongly supported by the ANOSIM result with a p-value of 1E-04 and an R-value of 0.7346 (Table S5). To pinpoint the specific genera responsible for driving this separation between early summer and late summer, a RandomForest analysis was performed. The analysis reveals several key genera that contributed significantly to the observed differentiation (Fig. 2g, h). These genera included C39, Candidatus Alysiosphaera, Dysgonomonas, Roseomonas, and Taibaiella. The variation of these genera likely played a crucial role in shaping the distinct microbial compositions observed between the early summer and late summer clusters. Among these genera, except for Roseomonas, the other four were found to be more predominant in the late summer samples. It is worth noting that Taibaiella exhibited the most pronounced discrimination, showing a 23-fold higher relative abundance in the late summer samples. Furthermore, it is noteworthy that certain species within the Roseomonas genus, e.g., Roseomonas mucosa have been known to be opportunistic pathogens for humans (Dé et al. 2004). Their increased relative abundance in early summer samples might suggest a potential health risk during that time. Additionally, Dysgonomonas bacteria have been recognized as causative agents of gastroenteritis, particularly in immunocompromised individuals (Ryan and Sherris n.d.).

Upon examining each of the three clusters in the PCA plot in Fig. 2c, it is apparent that each cluster encompassed all samples from the seven departments, reflecting the seasonal impact witnessed in each department. However, the ANOSIM results in Table S5 presented contrasting findings, indicating that seasonal variation significantly influenced the bacterial compositions solely in the blood test room and emergency department. Conversely, for the remaining five departments, the ANOSIM outcomes suggest that bacterial compositions varied irrespective of seasonal changes. Upon further investigation into the reason behind these contradictory results, it appears that the inadequate sample size for the ANOSIM test may have played a role.

Overall, the observed alpha- and beta-diversity patterns underscore that the bacterial community structure in the hospital indoor environments was significantly influenced by seasonal variations. However, while diversity is generally considered beneficial for ecosystem health, it can also introduce challenges, particularly in healthcare settings where cross-infection is a concern for immunocompromised patients. As a result, understanding diversity dynamics can provide valuable insights into the ecological processes and environmental factors that shape microbial communities (Brown et al. 2004), contributing to a broader understanding of microbial ecology and ecosystem functioning.

Tracing the sources of bacteria in the hospital indoor environments

Regarding the common genus types present, the Venn analysis reveals several noteworthy findings in the hospital indoor environments. Across all seven departments, there were a total of 197 common genus types, indicating the presence of a core bacterial community within the hospital environment (Figure S3a). To illuminate such phenomenon, the activities of healthcare occupants had a significant influence on it within healthcare facilities (Hospodsky et al. 2012). Through their interactions with patients and contact with various surfaces, the occupants could unintentionally facilitate the transfer of bacteria from one patient to another, thereby influencing the spread of healthcare-associated infections. In addition to the common community types among the department, the common bacteria between water and air dust samples was also observed, with 201 common genus types being observed between these two mediums. The intimation is that bacteria sharing occurred frequently in water and air mediums. Specifically, bacteria present in water sources could become aerosolized and subsequently released into the air. This was particularly notable during the summer season when air conditioners were in operation and were cooled using water from the hospital water, which enormously increased the likelihood of bacterial exposure and potential transmission to individuals within the healthcare facility (Leung et al. 2019).

Assuredly, there is a well-established fact that in the outpatient setting, the blood test room functions as a focal point for accommodating patients from diverse origins. As a consequence, to further seek the sources of the genera present in the blood test room, a source-tracking analysis was undertaken. Indeed, the validation that demonstrates the reliability of selecting the “real” sink as the designated sink role in the SourceTracker2 analysis has been conducted by Wu et al. (2022a, b). Based on this evaluation, during the analysis, the blood test room was considered the “sink” for the communities (genus level), while the other six departments were considered the “sources.” The results of the analysis (Figure S3b) reveal that 46.8% of the genera in the blood test room originated from the other six departments, of which the gynecology and emergency departments contributed the most with 21.6% and 8.81% respectively. However, there were still 53.2% of the sources that could not be identified, indicating a substantial portion of the genera in the blood test room were outdoor origins.

Broad-spectrum profile of ARGs and MGEsMDR and MGE genes were the most abundant in the hospital indoor environments

The analysis of ARGs and MGEs reveals the detection of 107 target ARGs and MGEs across all samples. The total relative copy number per 16S rRNA gene copy of these ARGs and MGE genes was 74.1, with an average relative copy number of 0.0290. Figure S4a illustrates the relative abundance of the 107 ARGs and MGEs belonging to 14 different types in all the samples. It is witnessed that the largest proportion of detected genes were dominated by MDR and MGE genes, with a total relative copy number of 24.8 and 20.6, respectively. The MDR genes were further classified into the mobile genes and the genes that were not featured with the mobility with a relative copy number of 5.55 and 19.3, respectively. The MGEs category included integrases, transposases, and plasmids. Integrases had a relative copy number of 11.4, transposases had a relative copy number of 9.14, and plasmids that were represented by tra genes had a relative copy number of 0.0600. This indicates a high likelihood of horizontal gene transfer (HGT) risk within the hospital, as MGEs play a significant role in facilitating the transfer of ARGs. Following the MGE genes, ARGs delivering resistance to sulfonamide, MLSB, and aminoglycoside were detected, with 9.28 relative copies, 6.90 relative copies, and 4.13 relative copies, respectively. Noticeably, among beta-lactam-resistant genes, there were 57.0% (1.88 relative copies) extended-spectrum beta-lactamases (ESBLs), which has been defined as transmissible β-lactamases that can be exchanged between bacteria (Shaikh et al. 2015). These findings align with a study conducted by Li et al. (2019) on hospital samples collected from air-conditioning units, which also identified similar patterns of dominating drug resistance.

Heterogenous ARGs/MGEs abundance and diversity in the hospital indoor environments

The different departments in the hospital indoor environments presented a heterogeneous distribution regarding the richness of ARGs and MGE genes (Fig. 3a). The blood test room had the highest abundance of ARGs and MGEs, with 12.9 relative copies, while the otolaryngology department had the lowest gene copies, with 6.18 relative copies. The distribution of MGE genes exhibited a different pattern (Fig. 3b), with the emergency department showing the highest abundance of MGE genes at 4.00 relative copies, indicating a higher frequency of gene transfer events in that department. In contrast to the relatively moderate level of heterogeneity in terms of ARGs and MGEs abundance, the diversity of ARGs and MGEs displayed pronounced fluctuations across departments. The surgery samples exhibited a notable increase in diversity, with an average Shannon index of 2.26 (Fig. 3c). This could be attributed to factors such as the use of antimicrobial chemicals to treat wounds and injuries (Hartmann et al. 2016), and the contribution of human sources from the high occupancy rates (Prussin and Marr 2015).

Fig. 3

a Abundance of ARGs and MGEs genes in each department, where the abundance represents the relative copy number per 16S rRNA gene copy. The upper edge of the error bar represents the abundance value plus the standard deviation (SD), and the lower edge of the error bar represents the abundance value minus the SD. The five-star symbols in the bottom of the bar in the emergency, surgery, pneumology, and otolaryngology departments represent a significant correlation between ARGs and MGEs in each of those department. b Abundance of MGEs genes in each department, where the abundance represents the relative copy number per 16S rRNA gene copy. c Boxplot of ARGs/MGEs Shannon diversity in each department, where the inside line in the each bar represents the median value, and the outside circles stand for outliers

Given the intricate and diverse distribution of ARGs and MGEs in hospital indoor settings, the primary factors shaping their compositions were expected to be complex. In detail, despite seasonal conditions like temperature and humidity exerting a significant influence on communities’ composition, both PCA and ANOSIM analyses indicate that seasonal variations were not the primary drivers altering the composition of ARGs and MGE genes (Figure S5a, Table S6), even though a clear seasonal pattern emerged with higher relative gene copies (48.4) and Shannon diversity (1.97) of ARGs and MGE genes during the summer months (Figure S6a-b). Actually, while a study by Lu and Lu (2020) stated that in the estuarine sediment environment, seasonal changes had a great impact on the abundance of ARGs, and there remains a dearth of research demonstrating the seasonal influence on shaping ARG compositions in hospital environments. Similarly, neither medium nor department variations significantly affected the components of ARGs and MGEs as well (Figure S5b-c, Table S6). While there may be differences in the richness, diversity, and specific types of ARGs/MGEs between different environmental factors (Figure S6), these differences may not be strong enough to be statistically significant. Other factors, such as microbial community dynamics, host factors, or specific environmental conditions, could also contribute to the observed patterns.

Department-specific correlation between ARGs and MGEs

It has been stated that MGEs can contribute to the HGT of ARGs among different microbes, and the host range of the MGEs carrying an ARG is important for determining how far it will spread (Ben Maamar et al. 2020). Consequently, to further prove the mobility of ARGs, the Mantel test between ARGs and MGEs was conducted. The results reveal a moderate correlation between the abundance of ARGs and MGE genes (p = 0.001, r = 0.2100) in the overall hospital indoor environment. This suggests that the abundance of ARGs was likely to be associated with the presence of MGEs, further supporting the notion of horizontal transfer of ARGs among bacteria in the hospital. Nevertheless, when the Mantel test was performed in each department individually, an intriguing finding emerged. The significant correlation between ARGs and MGEs was observed simply in the emergency, surgery, pneumology, and otolaryngology departments (Fig. 3a), while no significant correlation was found in the other three departments (Table S7). In the emergency, surgery, pneumology, and otolaryngology departments, there were specifically 44, 68, 59, and 45 pairs of ARGs and MGE genes, respectively, showing the robust correlation (p < 0.5, ρ > 0.6) based on the co-occurrence analysis. Consequently, to assess whether those ARGs correlating with MGEs exhibited greater similarity between seasons compared to those not correlated with MGEs within these departments, more ANOSIM tests were conducted for correlated ARGs between seasons and non-correlated ARGs between seasons. The results of these ANOSIM tests (shown in the Table S8) imply that there were no significant differences in ARGs composition between seasons within each department. This suggests that the similarity in ARGs composition between seasons was not primarily influenced by the transfer of ARGs carried by MGEs. Other elements, such as microbial structure (He et al. 2020), antibiotic usage patterns (Perry et al. 2021), persistent sources of ARGs, and indoor conditions (Perry et al. 2021), may have contributed to the observed similarity in ARGs composition between seasons in the hospital indoor environment.

Based on the ARGs’ mobility indicated by the significant correlation between ARGs and MGEs, the phenomenon of common ARGs and MGEs in the hospital indoor environments is manifested in Fig. 4a. Specifically, the indoor hospital exhibited a core set of 38 ARGs and 8 MGE genes across the studied departments. Significantly, akin to the unique trends observed in representative ARGs, the singular presence of specific ARGs was apparent. For example, the ophthalmology department had vancomycin-resistant genes vanSB and vanWB. The pneumology department had peptide-resistant gene mcr-2 and MLSB-resistant gene ermY. Additionally, the otolaryngology department had the aminoglycoside-resistant gene armA. Among those unique ARGs, vanSB, vanWB, and ermY have been acknowledged as chromosome-mediated ARGs (Leigh et al. 2022); consequently, their unique presence in specific departments could be attributed to their inherent mobility. In terms of mcr-2 and armA gens, they have been known as mobilized resistance genes mediated by plasmids (Liassine et al. 2016; Saadatian Farivar et al. 2018). Hence, the presence of their unique existence in specific departments might result from their limited transferability mediated by MGEs.

Fig. 4

a Venn diagram of common ARGs and MGE genes in seven departments. b Source-tracking analysis of ARGs and MGEs across seven departments, where the blood department acted as “sink”, and other six departments served as “sources”

Once more, a source-tracking analysis was performed to probe the origins’ proportions of ARGs and MGEs found in the blood test room. To the same token, the analysis maintained the postulation that the blood test room functioned as a “sink” for the accumulation of ARGs/MGEs, with the other six departments acting as the “sources.” According to the source tracking results (Fig. 4b), it was found that 50.8% of ARGs and MGEs present in the blood test room originated from the other six departments. Among these, the gynecology department was identified as a significant contributor, accounting for 31.5% of the ARGs and MGEs, while the pneumology and surgery departments contributed 7.90% and 5.45% respectively. Considering the gynecology department as one of the sources that contributed 21.6% of the genera to the blood test room, it suggests that the gynecology department played a non-negligible role in the dissemination of bacteria, ARGs, and MGEs. However, it is crucial to note that a substantial portion (49.2%) of the source of ARGs/MGEs could not be identified, implying the possibility of outdoor sources. A similar source-tracking study was also conducted by Li et al. (2021) in the same hospital, which pointed out that in the air dust, the outpatient hall was one of the main ARG transmission sources to the ophthalmology and pediatrics departments.

Co-occurrence between bacterial communities and ARGs/MGEs in the gynecology department

In this study, ARGs were able to be found in all studied hospital indoor environments. Indeed, the use of antibiotics by humans in the hospital has been found to stimulate the acquisition of ARGs by pathogenic bacteria (Martinez et al. 2009). To investigate whether the bacterial community was correlated with ARGs and MGEs composition, we used the Procrustes analysis and the Mantel test to correlate profiles. The Procrustes test shows that there was not a goodness-of-fit test (Table S9, M2 = 0.6583, p = 1E-06, 10,000 permutations) on the basis of Bray − Curtis dissimilarity metrics, reflecting the potential inconsistency between the bacterial community composition and the composition of ARGs and MGEs. The mantel test proved this as well (Table S9, r = 0.03715, p = 0.113). However, in contrast to the insignificant relationship between all detected genera and ARGs/MGEs, the pairwise Spearman’s rank correlation (23,221 pairs in total) displayed the intriguing result that tetracycline-resistant gene tet(36) was significantly connected with nine genera (p-value < 0.01, ρ > 0.6). It has been previously reported that tet(36) was first discovered in Bacteroides sp. strain and HGT of tet(36) was claimed to occur frequently between divergent phylogenetic groups in the farm environment (Whittle et al. 2003a, b). Further, to assess the relationship between genus types of ESKAPE pathogens and ARGs/MGEs, another co-occurrence analysis was conducted. Specifically, the analysis focused on the genera of Pseudomonas, Enterobacter, Staphylococcus, Klebsiella, and Acinetobacter, in relation to 107 ARGs/MGE genes. The results of the Procrustes test indicate a lack of significant fit (Table S9, M2 = 0.7946, p = 1E-04, 10,000 permutations). This inconsistency was further supported by pairwise Spearman’s rank correlation analysis. Among the total of 540 pairs examined, none exhibited a p-value less than 0.05 and ρ greater than 0.6, reinforcing the absence of a strong correlation between the two factors.

Absorbingly, although there was no correlation observed between the community composition (even ESKAPE) and ARGs/MGEs profile in the overall studied environment, a strong connection between them was discovered specifically in the gynecology department which was supported by the Procrustes’residuals value. In general, the residuals value represents the closeness between the ARGs/MGEs and genus composition. A lower residuals value indicates a closer relationship. In the gynecology department, we found that the average residuals value of samples was significantly low, suggesting a strong association between ARGs/MGEs and bacterial composition in that specific department. To further evaluate this strong correlation, Procrustes and Mantel analyses were conducted exclusively targeting the ARGs/MGEs and bacterial composition in the gynecology department. Both analyses gave compatible results, showing a strong correlation between ARGs/MGEs and bacterial composition (Table S9, Procrustes with M2 of 0.1007 and p of 0.0025; Mantel test with r of 0.7571 and p of 0.018).

Based on these results, we proceeded to perform pairwise Spearman’s correlation analysis between nine gynecology samples containing 43 ARGs/MGEs, and 181 genera. The results show that all 43 target genes and 181 genera had a significant relationship, with p values below 0.05 and \(\rho\) values above 0.6. Moreover, 32 ARGs and 117 genera exhibited \(\rho\) values above 0.8, with p values below 0.01. In the study conducted by Li et al. (2015), it has been hypothesized that the non-random co-occurrence patterns between ARGs and microbial taxa could indicate the possible host information of AGRs if the ARGs and the co-existed microbial taxa possessed significantly similar abundance trends among the different environments (Spearman’s ρ > 0.8, p-value < 0.01). In line with this hypothesis, those matched correlation values of 32 ARGs/MGEs and 117 genera were sent for the undirected network analysis.

The created network of the gynecology department consisted of 185 nodes (each node represented a subtype of ARGs, MGEs, or bacterial taxa) and 366 edges (Fig. 5). The modularity index was 0.642 (value > 0.4), suggesting that the network had a modular structure. Based on the modularity class, the entire network was separated into three major modules. Compared with a random association, clusters of nodes in the same module contained more interactions among themselves than with other nodes. Module I was the largest module comprising 82 nodes, followed by modules II and III including 23 and 20 nodes, respectively. From the network, it was witnessed that transposase Tp614 and aminoglycoside-resistance sat4 in module I had the most positive connections with 151 types of genera individually. Besides Tp614, MGE genes including integrase int1-a-marko and transposase tnpA 201 in module III had more positive connections with 18 and 17 types of genera as well. Consequently, the substantial number of connections involving MGEs in the network indicate the dissemination of ARGs and the potential acquisition of these ARGs by other microbes within the gynecology department. This phenomenon underscores the impact of a high-antibiotic-selective pressure in the hospital environment. Other ARGs, such as bacA, ampC in module III, and tetX in module V, which conferred resistance to bacitracin, beta-lactamase, and tetracycline, respectively, were associated with multiple candidate genera (20, 15, and 12 genera, respectively), suggesting a higher likelihood of these ARGs being carried by a diverse range of microbial hosts. However, it is important to note that while many ARGs were associated with multiple genera, only a few ARGs were found to be potentially carried by a single host. For example, the MLSB-resistant gene msr(E) was exclusively correlated with Chelatococcus, and Paludibacter was the only connection with MLSB-resistant gene lnuA as well. These observations indicate that certain ARGs may have a more restricted distribution and are less likely to be transferred between different microbial taxa.

Fig. 5

a Co-occurrence analysis between bacterial communities (genus-level) and ARGs/MGEs in gynecology department with nine samples. Circle colors represent ARGs and MGEs types, and dark-green boxes represent the genera. Line colors mean positive (red) or negative (blue) spearman correlation. Font/node size represents the degree of connections. Font color with red represents the concerned genera. Line width represents the absolute value of the correlation index, i.e. the wider the line shows, the bigger the absolute value is. b the same network as Fig.(a), while the node colors here represent the Module classes

Furthermore, the network analysis unveiled significant correlations between certain clinically relevant genera and specific ARGs/MGEs. Notably, Legionella exhibited a significant correlation with two ARGs associated with antibiotic resistance. Specifically, it was found to be linked to the beta-lactam-resistance gene blaOXA10 and the MLSB-resistance gene va0. Another notable discovery affirmed a correlation between Enterobacter (ESKAPE-related) and the MDR cefa_qacelta, pinpointing the association between this genus and multidrug resistance. What is more, Acinetobacter (ESKAPE-related) was acknowledged to be co-occurring with vancomycin-resistant vanB, of which the presence has been known to signify a heightened level of resistance to vancomycin (Luqman et al. 2024). Of particular concern were the correlations observed between several genera, including Bacillus, Escherichia-Shigella, Klebsiella, Neisseria, Staphylococcus, Stenotrophomonas, Streptococcus, and sat4/Tp614. Noticeably, the network presents that Sandaracinobacter and Azorhizobium had a close loop cycle with blaPSE. This implicates that blaPSE was the low-risk ARG that might not transfer in the gynecology department.

In summary, the network analysis reveals the modular structure of the gynecology department’s bacterial community, with specific ARGs, MGE genes, and bacterial taxa–forming distinct modules. MGEs played a prominent role in facilitating the dissemination of ARGs, indicating the potential for HGT among different microbial organisms. While the majority of ARGs showed correlations with a wide range of microbes, others such as blaPSE were exclusively associated with specific bacterial taxa, highlighting the immobility and conservation of certain ARGs.